Gemini 3 Deep Think: Optimizing 2D Semiconductor Fabrication

TL;DR: Leveraging DeepMind's "Deep Tank" AI system, the lab successfully grew a 2D semiconductor measuring 130 micrometers, surpassing the 100-micrometer target and paving the way to overcome the limits of silicon-based materials. ## Challenges and Opportunities of 2D Semiconductors ### Why 2D Materials Are Critical Two-dimensional (2D) materials are those with a thickness of just a single molecule. Their extreme thinness makes them an ideal candidate for future electronic devices. As silicon approaches its physical limits, exploring new materials in the 2D domain has become a key direction. ### The Difficulty of Growing 2D Semiconductors Growing 2D materials is highly challenging, with the core difficulty lying in parameter selection: - Gas flow rates must be precisely adjusted - Furnace heating must be controlled simultaneously - Experts typically require weeks or even months to find the optimal parameters ## Breakthroughs by Deep Tank ### AI-Recommended Recipes Exceed Expectations In my lab, we use Deep Tank to design new semiconductors. The goal was to grow a 2D semiconductor of 100 micrometers. By using the recipe recommended by Deep Tank, we achieved a size of 130 micrometers—the best result the lab has ever seen. ### From a Single Value to a Full Temperature Curve Deep Tank does not just provide a single temperature value; it outputs a complete temperature curve. It systematically integrates the latest advances in science, significantly shortening the parameter optimization process that previously took months. ## Future Outlook ### New Possibilities with the Deep Sync API "I am very excited; this is just the beginning." The Deep Sync API opens new doors for automating many existing instruments, meaning AI can not only optimize process recipes but also directly control experimental equipment, enabling closed-loop automation from design to execution. ### Towards Automated Science As silicon approaches its limits, my lab is using Deep Think to explore new materials in the 2D domain. This AI-driven experimental paradigm promises to accelerate the entire workflow from material discovery to device manufacturing. --- Source: Gemini 3 Deep Think: Optimizing 2D semiconductor fabrication - Google DeepMind (https://www.youtube.com/watch?v=IE0BmXsIzTI)